Mandik repeatedly refers to his Unicorn Argument as “the Unicorn,” creating sentences such as, “In sections 4 and 7, I examine and reject proposals that HORs and FORs may save themselves from the Unicorn by embracing the Direct Reference hypothesis (DR).”
The Unicorn is coming! Save yourself!
Archive for the ‘Unicorn Argument’ Category
My paper, “Beware of the Unicorn: Consciousness as Being Represented and Other Things that Don’t Exist” just came out in the latest issue of Journal of Consciousness Studies. [link to uncorrected proofs] [link to IngentaConnect] What I didn’t know until I got my hands on the issue, is that my unicorn made the front cover. I can’t wait to tell my mom. Enjoy, but beware, this jpeg:
Longtime Brain Hammer readers, if any are left after the long hiatus, may recall the many past and heated discussions of my “Beware of the Unicorn: Consciousness as Being Represented and Other Things that Don’t Exist”. What was once a blog discussion is now an article. It’s forthcoming in Journal of Consciousness Studies. 16(1). [link to draft file].
Abstract: Higher-Order Representational theories of consciousness — HORs — primarily seek to explain a mental state’s being conscious in terms of the mental state’s being represented by another mental state. First-Order Representational theories of consciousness — FORs — primarily seek to explain a property’s being phenomenal in terms of the property being represented in experience. Despite differences in both explanans and explananda, HORs and FORs share a reliance on there being such a property as being represented. In this paper I develop an argument — the Unicorn Argument — against both HORs and FORs. The core of the Unicorn is that since there are mental representations of things that do not exist, there cannot be any such property as being represented, and thus no such property with which to identify either being conscious or being phenomenal.
Pete Mandik, Sick, 2006
Chapter 3 of The Subjective Brain is up now, and it’s called Beware the Unicorn: Consciousness, Intentionality, and Inexistence.
In case you think you’ve heard it (and hated it) all before, pause and appreciate this: you haven’t. There’s new stuff peppered throughout. See, for instance, the brand-spanking-new section 7. Enjoy!
The conclusion of the Unicorn argument is incompatible with HOT and FOR. HOT and FOR derive much of their plausibility from Transitivity and Transparency, respectively. If the lesson of the Unicorn is something that we can live with, then perhaps we must either (1) learn to live without Transitivity and Transparency or (2) find a way of accepting Transitivity and Transparency while rejecting HOT and FOR. Option (1) is the best option. Option (2) is unwelcome because it is hard to see how Transitivity and Transparency donâ€™t just lead relatively directly to HOT and FOR, respectively. Further, a direct case for (1) can be made, and it is the aim of this section to make it. Resistance to abandonment of Transitivity and Transparency may be due to the fact that both theses are prima facie plausible and arguably useful. However, I think that their plausibility can be explained away and their utility can be had by much more plausible substitutes.
My case against Transitivity will have three parts: (1) its plausibility can be explained away, that is, its plausibility can be explained without supposing it true, (2) if Transitivity is supposed to be analytic, then a certain situation which is not obviously incoherent would be obviously incoherent, and (3) if Transitivity is not supposed to be analytic, but instead defended on grounds of theoretical utility, then it may just as well be replaced by what Iâ€™ll call Deflationary Transitivity.
Origami unicorn from Blade Runner. A non-existent representation of a non-existent.
Following philosophers like Tim Crane and Uriah Kriegel, letâ€™s call the Problem of Intentionality the problem of motivating the rejection of one of the three propositions in the following inconsistent triad:
1. We think about non-existents
2. One can bear relations only to existents
3. Thinking about is a relation
Part of my interest in the Problem of Intentionality is that a big chunk of the Unicorn Argument involves an acceptance of 1 & 2 and a rejection of 3.
Iâ€™ve gotten grief from philosophers like Chase Wrenn and Eric Steinhart about whether the Unicorn can be stated in a formal calculus. Such grief can equally be directed at the Problem of Intentionality. We can motivate such grief by formulating what Iâ€™ll call the Steinhart Principle:
Steinhart Principle: A set of propositions exhibits logical properties (e.g., validity, inconsistency) only if there is at least one calculus in which the propositions are jointly formalizable.
I have a worry about the applicability of the Steinhart Principle to either the Unicorn or Intentionality that I would like to raise in terms of what Iâ€™ll call the Mandik Principle:
Mandik Principle: The adoption of a formalism is philosophically fruitful only if doing so doesnâ€™t beg (pro or con) the question at hand.
Consider, then, the following challenge: State the Problem of Intentionality in a way that simultaneously respects both the Steinhart Principle and the Mandik Principle.
Can this challenge be met? I havenâ€™t made up my mind one way or another, but here are some reasons for doubting that the challenge can be met.
Consider that meeting this challenge would involve formulating the three propositions in a way that doesnâ€™t require one to assign a particular truth-value to any of them. Now consider proposition #1. It is very difficult to see how to proceed with its formalization without also taking a stand on the truth of 1, 2, or 3. For exampleâ€¦
Suppose that we formulate 1 as
($x)($y)(Px & ~Ey & Txy)
where â€œ($x)â€ is the existential quantifier, â€œPxâ€ is â€œx is a personâ€, â€œExâ€ is â€œx existsâ€, and â€œTxyâ€ is â€œx thinks about yâ€.
Lots of problems arise aside from the fact that one may be squeamish about an existence predicate. In particular, formulating 1 in terms of the two-place â€œTxyâ€ presumes the truth of proposition 3.
On the other hand, we might try to formulate 1 as
($x)($y)[Px & Tx & ~($z)(Uz)]
where â€œUxâ€ is â€œx is a unicornâ€ and â€œTxâ€ is a predicate we construct by presuming a language of thought and an apparatus of thought-quotation giving us â€œx is thinking â€˜($z)(Uz)â€™â€.
On this formulation lots of problems arise aside from the fact that we are quantifying into the opaque context of thought quotation. In particular, it looks like such a formulation in terms of a one-place thinks predicate presumes the falsity of 3.
Letâ€™s suppose for the sake of conversation that there is no formalization of the Problem of Intentionality that satisfies the Mandik Principle. What, then, is the most appropriate response to the Problem of Intentionality? Rejecting it as a non-problem seems itself to beg genuine philosophical questions.
(1) Kryptonite is (numerically identical to) the mineral “sodium lithium boron silicate hydroxide” [according to the label shown in the film Superman Returns]
(2) Kryptonite is essentially fictional
(3) Sodium lithium boron silicate hydroxide is actual (and so not essentially fictional)
Chappell argues for ditching (1), but my inclination is against (2). I figure that ficitonal entities don’t literally have any properties yet alone essential ones. As I argued in “Dear Watson” there might be an attenuated sense in which fictional entities have properties in virtue of authorial intent, but they will seldom have, in this sense, the property of being represented. For similar reasons, they won’t have the property of being fictional.
One worry I have about the specific example of Kryptonite is that there is too much divergence between the AP reported substance and the various properties attributed in the Superman stories. I hoped to get around this with a chemically pure (pun!) version of the puzzle: The Puppy Crusher. From the previous comments thread:
Suppose that in a James Bond novel, a character mixes a drink that no one has ever mixed before - say itâ€™s three parts gin and one part maple syrup - and they call it â€œThe Puppy Crusherâ€. Suppose at some later date an actual bar tender mixes up three parts gin and one part maple syrup. Is it necesarily true that that drink isnâ€™t a Puppy Crusher?
From the Associated Press:
New mineral found has same composition as fictional kryptonite
LONDON â€” A mineral recently discovered in Serbia has the same composition as kryptonite â€” the fictional substance that robs Superman of his powers â€” the British Museum said today.
While the material is not a perfect match, its chemical breakdown is strikingly similar.
A drill core of the unusual mineral was unearthed in Serbia by the mining group Rio Tinto PLC, which turned it over to mineral expert Chris Stanley at the Natural History Museum for analysis.
“Towards the end of my research I searched the Web using the mineral’s chemical formula, sodium lithium boron silicate hydroxide, and was amazed to discover that same scientific name written on a case of rock containing kryptonite stolen by Lex Luthor from a museum in the film Superman Returns,” Stanley said.
The material is white, powdery and not radioactive â€” unlike the glowing green crystals usually depicted in the Superman comics. It will be formally named Jadarite when it is described in the European Journal of Mineralogy later this year.
Approximately 30 to 40 new minerals are discovered each year, the museum said, although usually only in the form of a few grains only visible under the microscope.
From Kripke’s Naming and Necessity p.: 156
There were two theses: first, a metaphysical thesis that no counterfactual situation is properly describable as one in which there would have been unicorns; second, an epistemological thesis that an archeological discovery that there were animals with all the features attributed to unicorns in the appropriate myth would not in and of itself constitute proof that there were unicorns.
So…kryptonite remains undiscovered?
In “Me So ‘Corny“, I examined and rejected the proposal that maybe a kind of direct reference can save HOR (Higher-Order Representational) theories of consciousness such as HOT (Higher-Order Thought) from the Unicorn. I want to do a similar thing here for FORs (First-Order Representational theories). The proposal of uniting FOR with DR raises special issues. One issue is that FORs concern representations of properties, not particulars. The second issue is that FORs concern representation in experience, not thought.
Recall that, for HORs, DR was described as holding that there are certain mental representations such that (a) two or more of these representations are about the same object if and only if they have the same cognitive significance and (b) these representations have representational content only if that which they represent exists. The question arises: what is the most straightforward way of adapting DR to fit with a theory of the representation of properties in experience? I think that (a) and (b) can serve as useful models. We can attempt to make suitable alterations, (a+) and (b+). The transformation of (a) into (a+) will obviously involve replacing â€œobjectâ€ with â€œpropertyâ€. Not so obvious is what to do with â€œcognitive significanceâ€ although â€œexperiential significanceâ€ might suffice. Or, more to the point of a discussion of phenomenal consciousness, we may work with â€œwhat itâ€™s like,â€ where sameness and difference in experiential significance may be regarded as sameness and difference in what itâ€™s like. Thus we have
(a+): Two experiences represent the same property if and only if they are the same with respect to what itâ€™s like to have them.
Moving on to the modification of (b), the main problem to deal with is how to apply the exists/doesnâ€™t exist distinction to properties instead of objects. Two suggestions immediately arise. The first is to identify it with the instantiated/uninstantiated distinction. The second is to identify it with the possibly instantiated/ necessarily uninstantiated distinction. I will focus on the second option, since I intend to present counter examples to FOR+DR and counter-examples to FOR+DR in terms of necessarily uninstantiated properties are a fortiori counter-examples to FOR+DR in terms of uninstantiated properties. Thus, part of what is entailed by combining direct reference with FOR is
(b+): An experience represents a necessarily uninstantiated property if and only if there is nothing it is like to have the experience.
In what follows I will argue against the wedding of FOR and DR by arguing that there can be experiences for which there is something it is like but the represented property is necessarily uninstantiated.
We see (that is, visually represent) necessarily uninstantiated properties whenever we look at certain pieces of art by M. C. Escher. In many of Escherâ€™s artworks, we see what at first glance seem to be three dimensional objects and their arrangements, but on further reflection couldnâ€™t possibly exist. For example, in Escherâ€™s 1960 lithograph, â€œAscending and Descending,â€ we see (and thus visually represent) a finite set of stairs, each one of which is higher than some other.
Now, it is open for the FOR theorist to hold that what is paradoxical in viewing such a picture is restricted to what concepts one brings to bear on the experience and that the contents of the experiences themselves contain nothing contradictory because, for example, the contents of the experiences themselves concern only the representation of a distribution of shades of gray in the visual field. I donâ€™t think this response is particularly plausible, but I wonâ€™t pursue this further here, for I think there are bigger and much more interesting problems for the FOR theorists, problems that arise from experiences with paradoxical contents not obviously attributable to any coinciding conceptual states.
Consider, for one such example, experiences of the motion aftereffect, or, more colloquially, the waterfall illusion. The effect occurs when one has been staring at a moving stimulus for a while, such as a waterfall, and then directs oneâ€™s attention to a stationary object such as a rock wall. One will then undergo a paradoxical experience whereby one and the same object, the rock wall in this case, appears simultaneously to be moving and not moving.
The problem posed by the motion aftereffect is that it is a putative example in which the property experiencedâ€”the property of simultaneously moving and not movingâ€”cannot be instantiated, for nothing in reality can be simultaneously moving and not moving. At this point, the FOR theorist may be tempted to re-describe the experience in question as actually being two experiences, one of which is an experience of something as moving and the other of which is an experience of the very same thing as stationary. Such a move would block the attribution of representations in experience of necessarily uninstantiated properties. However, one might wonder what independent motivation can be provided for such a move so as to make it not so obviously ad hoc. Instead of dwelling further on the motion aftereffect, I would like to spend time on a class of examples even more powerful.
Due to peculiarities of the normal functioning of the visual system, we can experience colored after-images. Readers are no doubt aware that after staring at a bright red spot and then directing their gaze at a white wall, they will experience a green afterimage. FORs provide a natural explanation of such after-images: though no green object need be present in the room, one undergoes so-called green afterimages in virtue of mentally representing in experience the instantiation of green in a certain region in space.
Under certain conditions, there can be induced in normal subjects afterimages with colors corresponding to no color an object can have. Following Paul Churchland, let us call such colors â€œchimerical colorsâ€ for they are â€œcolor[s] that you will absolutely never encounter as an objective feature of a real physical object.â€
The textbook case of an afterimage involves locating the afterimage on a white background by fixating oneâ€™s gaze on a white wall or piece of paper. Chimerically colored afterimages may be achieved when afterimages are located on non-white and non-gray backgrounds. For example, if one were to look at a pale-blue-green stimulus and then position the resultant orange afterimage on a maximally saturated orange background, the resultant afterimage will be colored what Churchland calls â€œhyperbolic orangeâ€ an orange which is â€œmore â€˜ostentatiously orangeâ€™ than any (non-self-luminous) orange you have ever seen, or ever will see, as the objective color of a physical object.â€
Locating afterimages on black backgrounds yields afterimages that no objects, self-luminous or not, could have. If one looks at a saturated yellow stimulus for 20 seconds and positions the blue afterimage on a black background, the resultant afterimage will still be blue but will be exactly as dark as black. This is especially interesting since, as Churchland points out, â€œno objective hue can be as dark as that darkest possible black and yet fail to be black.â€ Even more interesting is what happens when one starts by looking at a saturated blue and positions a yellow afterimage on a black background. The resultant image is still yellow, but a yellow exactly as dark as black. This is especially interesting because we tend to think of yellow as a light hue. Ludwig Wittgenstein once asked â€œ[W]hy is there no such thing as blackish yellow?â€ The afterimages described by Churchland show that while there cannot be such a thing as blackish yellow, it may nonetheless be represented in experience. The representation of blackish yellow involves the representation of a necessarily uninstantitated color, and as such, cannot be accommodated by any version of FOR wedded to DR.
Recall the sorts of objections the imagined FOR theorist raised against the Escher and waterfall illusion counterexamples to FOR+DR and note how ineffective such objections would be against the case of chimerically colored afterimages. The objection against the Escher case was that the paradoxical contents were represented in conception, not experience. Whatever plausibility such an objection had in the case of viewing a picture of an ever-ascending staircase, it certainly has no plausibility in the case of colored afterimages. The objection against the waterfall illusion was that perhaps what was happening was not a single experience of motion and its negation, but two distinct experiences, one of motion, and one of the lack thereof. Whatever plausibility such an objection had in the case of the waterfall illusion, it certainly has no plausibility in the case of colored afterimages. It is quite clear that when one as an experience of a color patch, even in the case of an afterimage, one is not undergoing three separate experiences, one each for the hue, the brightness, and the saturation of the color in question. One is, instead, having a single experience, one which involves the representation of a single color which, if instantiated, would also instantiate a particular hue, brightness, and saturation.
One possible FOR-friendly response would be to say that the necessarily uninstantiated properties described above are complexes of properties that are individually instantiable. Such a response would involve modifying FOR so that what it is like is solely determined by the atomic properties represented, not by their combination. But such a revision runs into a big problem, namely, that it makes binding irrelevant to what itâ€™s like. To see this point about binding, consider that thereâ€™s a difference in what it is like to see (1) red squares and blue circles and (2) blue circles and red squares. However, the possible response under examination would make (1) and (2) subjectively indistinguishable, for the response under examination would make the sole determinants of what itâ€™s like the representation of redness, blueness, square-ness, and circularity.
Over at the Philosophy of Brains blog, Richard Brown has a post called “Kripke, Consciousness, and the ‘Corn” in which he tries to defend Higher-Order Representational theories of consciousness against the Unicorn Argument, by wedding HOR to a Kripkean causal theory of reference.
By amazing coincidence, I have a new version of “Beware of the Unicorn” up on my website now which contains a section dealing with exactly this sort of move. I discuss this under the heading of combining HOR with a so-called “direct” theory of reference.
The direct reference hypothesis (DR) holds that there are certain mental representations such that (a) two or more of these representations are about the same object if and only if they have the same cognitive significance and (b) these representations have representational content only if that which they represent exists. The most promising aspect of DR with respect to defeating the Unicorn is (b). Combining HOR (Higher-Order Representational theory of consciousness) with DR entails postulating that all of the higher order representations relevant to explaining consciousness have representational content only if the states they are representations of exist. If HOR could be combined with DR it would be immune to the Unicorn.
However, at least in the case of HOT (Higher-Order Thought theory of consciousness), HOR cannot be plausibly combined with DR. This is due to troubles that arise in connection with part (a) of DR. When we examine the most plausible examples of attributions of consciousness-conferring higher-order thoughts, we find that they give rise to opaque contexts inconsistent with DR.
To see these points, consider an example. Suppose that Jones has some mental state that is a candidate for state consciousness. Suppose, then, that Jones sees that x is red. In order for the state of Jones seeing that x is red to be a conscious state, according to HOT, Jones must have a higher-order thought about that state. It is useful to consider what attributions of that thought would look like. We might attribute the HOTs by saying that
(1) Jones believes that he sees that x is red.
(2) Jones believes of himself that he sees that x is red
Either way, by the time we get to â€œhe sees that x is redâ€ we are well into an opaque context.
Suppose that seeing that x is red is identical to having neural activity pattern number 67 in area v4 of cerebral cortex. Consider that if we replace â€˜sees that x is redâ€™ in (1) and (2) with â€˜has activity 67 in area v4â€™ then we wind up with sentences that may very well have the opposite truth values of (1) and (2). This is not because seeing that x is red is not identical to having activity 67 in area v4. This is because Jones may very well lack appropriate neurophilosophical sophistication to believe of himself that he has activity 67 in area v4. DR requires the intersubstitutability salvae veritate of co-referring terms for the alleged relata. If the defender of transparent HOTs were to insist on the possibility of the above substitutions as salva veritate, then the following problem arises. If the meaning of a term is purely referential, and HOTs determine what it is like, and â€˜sees that x is redâ€™ and â€˜has activity 67 in area v4â€™ are co-referring, then Jonesâ€™s perceptual experiences would seem to him to be the neural activity pattern 67 in area v4. I suppose, however, that while Paul Churchland’s experiences may seem neural to Paul Churchland, Jonesâ€™s experiences need not seem neural to Jones.
Perhaps a different way of attempting to wed HOT and DR is by construing consciousness-conferring higher-order thoughts as referring demonstratively. Such a construal would entail that Jones’ state of seeing x as red is conscious only if Jones has a Higher Order demonstrative thought expressible by “this is a state of seeing that x is red” where the demonstrative “this” refers, if at all, to a state Jones actually has. If the demonstrative “this” fails to refer, then “this is a state of seeing that x is red” fails to express a consciousness-conferring higher order thought because it fails to express any thought.
One consequence of a direct reference theory of demonstrative thoughts is that any difference in reference of â€œthisâ€ gives rise to differences in thought. Two occasions of thoughts expressible by â€œthis is an umbrellaâ€ would be occasions of thoughts with different contents if the two occasions of the demonstrative â€œthisâ€ referred to numerically distinct umbrellas. We might summarize this point by saying that directly referring demonstrative thoughts are object-involving.
The object-involvement of demonstrative thoughts does not fit well with the HOT theory. The main problem arises because, on Rosenthalâ€™s HOT theory, the contents of HOTs are supposed to be responsible for determining what itâ€™s like to have conscious states.
Rosenthal states the relation between HOT and what it is like as follows:
What itâ€™s like for one to be in a qualitative state is a matter of how one is conscious of that state. If I am conscious of myself as having a sensation with the mental quality red, that will be what itâ€™s like for me, and similarly for every other mental quality. And how we are conscious of our qualitative states is a matter of how our HOTs characterize those states. There being something itâ€™s like for me to be in a state with a particular mental quality is a matter of my having a HOT that characterizes a state I am in as having that mental quality. (D. Rosenthal, Consciousness and Mind, p. 186.)
In other words, what itâ€™s like to be in a conscious state is one and the same as how oneâ€™s state appears to one. Further, how the state appears to one is a matter of how the state is represented by a higher-order thought.
The appearance-determining aspect of consciousness-conferring higher-order thoughts is the aspect that makes them so poorly modeled by demonstrative thoughts. Mere numerical differences can suffice to give rise to differences in demonstrative reference. However, mere numerical differences do not suffice to give rise to differences in appearance.
My physical doppelganger who lived on a physical doppelganger of the planet I live on, with a physically similar life history would, I take it, have conscious states such that what it is like for him to be in those states is like what it is like to be in mine. His object-involving thoughts, however, would differ from mine insofar as his â€˜thisâ€™â€™s pick out a distinct umbrella from mine, his â€˜hereâ€™â€™s distinct places, his â€˜Iâ€™â€™s a distinct person. But just as his umbrella may very well appear just as my umbrella does, so will his lower-order mental states appear to him as mine do to me. Thus, in spite of diverging in the contents of our demonstrative higher-order thoughts, what itâ€™s like to be me may very well be just like what itâ€™s like to be my physical doppelganger.